Summary: | 碩士 === 國立臺北科技大學 === 電子工程系 === 106 === Nowadays, the Adaptive Cruise Control (ACC) system has been equipped in many high-tech vehicles. By controlling the acceleration of the cars, ACC maintains adjustable distances which varies with the speed. However, the habit of the driver might not consist with the ACC system, and frequent adjusting may bother the drivers during driving. In this Thesis, a Personalized Adaptive Cruise Control (PACC) is proposed to learn the habit pattern and drive the car. In the learning mode, the PACC system can be trained by Fully Connected Deep Learning algorithm with filtered data. In the driving mode, the PACC system can determined the expected distance with detecting signals. The experimental results show that the parameter of Mean Absolute Percentage Error (MAPE) is 58.1% based on the proposed default mode. The MAPE is reduced 21.07% after training by the driver’s data.
|